Anomaly detection in biogas plants is critical for ensuring optimal performance and preventing potential failures. Traditional monitoring methods fail to capture complex sensor interactions, optimise reconstruction error and focus on critical features for accurate anomaly detection. This study develops a deep LSTM autoencoder with attention (DARE-LSTMAE) to enhance anomaly detection accuracy and enable real-time fault detection. It is designed to learn the normal operational patterns of the biogas plant by encoding and reconstructing the input data. The attention mechanism helps the model focus on the most important properties, improving anomaly detection. The experimental results demonstrate that the LSTM autoencoder with an attention mechanism outperforms traditional methods in detecting anomalies. The proposed model achieves 97% precision, 100% recall and 98% F1 score, providing early warnings of potential issues. The optimised reconstruction error further enhances the system's reliability, making it a robust solution for real-time monitoring and fault detection in biogas plants.
Meena et al. (Thu,) studied this question.